OMOP Data Linkage with dbt [WIP]

Overview

This work-in-progress dbt package is designed to facilitate the linkage of OMOP CDM databases from different sources—specifically EHR (Electronic Health Records) and Claims databases. Currently, the package supports PostgreSQL with 2 incoming data sources. The plan is to extend support to more data sources. The package provides a series of SQL models to:

• Match persons across databases based on person_source_value
• Pool other tables (e.g., visit_occurrence, condition_occurrence, observation_period) from OMOP CDM with updated person_ids
• Validate and identify discrepancies in key attributes like gender and year of birth
• Identify temporal discrepancies in condition_occurrence

Getting Started

Pre-installation Steps

1. Create New PostgresDB for Linked Data: Initialize a new PostgreSQL database where the linked data will reside.
2. Set Up FDW (Foreign Data Wrapper): Configure FDW for the two incoming data sources (EHR and Claims) to enable cross-database queries.

You may leverage this DDL template to create the FDW for each incoming data source:

CREATE EXTENSION IF NOT EXISTS postgres_fdw;

CREATE SERVER omop_ehr FOREIGN DATA WRAPPER postgres_fdw OPTIONS (host '{hostname}', port '{port}', dbname '{dbname}');

CREATE USER MAPPING FOR {host_user} SERVER omop_ehr OPTIONS (user '{foreign_user}', password '{password}');

CREATE SCHEMA IF NOT EXISTS cdm_ehr;

IMPORT FOREIGN SCHEMA {foreign_schema} FROM SERVER omop_ehr INTO cdm_ehr;

3. Run OMOP DDL for the New Linked Schema: Execute the DDL scripts to create the necessary OMOP tables in the new linked database.

Installation

4. Install dbt: If you haven't already, install dbt. Make sure to install the Postgres-compatible version (dbt-postgres).
5. Clone this repository: Clone this dbt package repository into your local machine.

Configuration

6. Configure your profiles.yml: Update your dbt profiles to point to your Postgres instances where the OMOP databases reside.
7. Configure models/sources.yml: Edit this file to specify where the incoming data sources (EHR and Claims) are located.

Models

Here is a list of the main models in this package:

Linkage Models

• linked_person.sql: Matches persons from EHR and Claims databases based on person_source_value.

Validation and Discrepancy Models

• person_discrepancy.sql: Identifies discrepancies in gender_concept_id and year_of_birth in the person table.
• temporal_discrepancy_condition_occurrence.sql: Identifies temporal discrepancies in the condition_occurrence table.

Summary Models

• person_summary.sql: Provides a summary of the number of matched, unmatched, and total persons in each source.

Running the Package

To build the models, navigate to the directory containing the dbt_project.yml file and run:

dbt run

Contributing

If you encounter any issues or have suggestions for improvements, please open an issue or submit a pull request.